Implement b-tag hypo for large-R jet tagging
This MR implements b-tagging hypo for large-R jet b-tagging (X to bb) and adds a test chain to Dev menu. The b-tagging is done using offline X to bb tagger (GN2Xv01) on a10sd cssk pflow chain only (one jet collection). The jets are tagged directly without creating a BTagging object. As a result, the case of large-R jet tagging has to be implemented in the Jet hypo, specifically scenario_simple.py
. For now there is only one working point; the working point was found by averaging the 60th percentile values for RS graviton to HH to 4b signal samples (M=300 to 1000 GeV). Looser working points can be added if needed. The tagger needs to be optimized for trigger jets in the future.
CC: @lshi and @lbeemste
Merge request reports
Activity
- Resolved by Roshan Joshi
Ugh, sorry for being so slow on this.
The problem with this is that it's implementing the cuts in the b-jet hypo, which is only being run for the b-jets we copy over from pflow jets after some selection is applied. It also requires that a
BTagging
object exist, which isn't true for large-R jets.I think the easier way to implement this is going to be to in the jet hypo, see for example the MR where we added fastDips (!49919 (merged)), or the one where Thandi tried to modify that (!63500 (closed)).
It might also help to look at what Stefano had to do to get b-tagging on pflow jets in !52412 (merged). We didn't implement a hypo in that case but we did have to add b-tagging to a new jet collection.
added 180 commits
-
b57f47e4...9f6a2075 - 180 commits from branch
atlas:23.0
-
b57f47e4...9f6a2075 - 180 commits from branch
added 3337 commits
-
9a42058f...b145d274 - 3336 commits from branch
atlas:main
- ee377905 - Try with main branch
-
9a42058f...b145d274 - 3336 commits from branch
Hi @dguest, now that I see some events passing the trigger (but using RS graviton to HH to 4b RDO), what should the next step be? I think the default tests run over ttbar, so the reference count file would have 0 counts for the new test chain. I would be happy to provide the RDO I created with 500 events if we want to have a mechanism to test with this RDO.
With the current setup (HH to 4b RDO), I have the following diff for reference count after running test for RDO_Trig from RDO:
+HLT_j100_a10sd_cssk_60bgntwox_pf_jes_ftf_L1J100: + eventCount: 17 + stepCounts: + 0: 146 + 1: 17 + stepFeatures: + 0: 146 + 1: 17
Note that the number of events I ran over was 150. With the default ttbar RDO, as mentioned, the counts are 0 for the b-tagging step:
+HLT_j100_a10sd_cssk_60bgntwox_pf_jes_ftf_L1J100: + eventCount: 0 + stepCounts: + 0: 3 + stepFeatures: + 0: 3
Currently, the "threshold" for b-tagging is set to the approximate average of the 60% quartile of the discriminant histogram of the signal samples that we have (M=300 to 1000 GeV). Not sure if that is the right way to go about setting threshold.
Hi @dguest, if you have time, could you please take a look at this? I tried the minimal runTrigART test. The RDO to RDOTrigger test has count change only for the new chain, as mentioned in the diff above. The
test_trig_data_v1Dev_build
test has warning message in CheckLog:xAOD::AuxSelection WARNING Selected dynamic Aux atribute "TrackWidthPt1000GN2Xv01_phbb" not found in the registry FAILURE : error/fatal found in athena.log
It says this is fatal but in athena.log, there are no instances of ERROR or FATAL that I could find. Same warning in CheckLog of
test_trig_data_newJO_build
.For
test_trigP1_v1Dev_decodeBS_build
test, the counts changed for many chains (1 less than reference for every chain). This is surprising to me since the added code should not affect any other chains.I don't know if something changed recently and the references were not updated. For reference I was using Athena 2023-09-14T2101 for the tests.Thanks!
added 1 commit
- 90faf770 - Correct the aux variables to be added to the large-R jet
added BJetTrigger JetHLTHypo labels
Jenkins please retry build.
Edited by Roshan JoshiThis merge request affects 5 packages:
- Trigger/TrigHypothesis/TrigHLTJetHypo
- Trigger/TrigSteer/TrigOutputHandling
- Trigger/TriggerCommon/TrigEDMConfig
- Trigger/TriggerCommon/TriggerJobOpts
- Trigger/TriggerCommon/TriggerMenuMT
Affected files list will not be printed in this case
Adding @martindl ,@fpastore ,@miochoa ,@carquin ,@ayana ,@bernius ,@okuprash ,@malconad ,@fwinkl ,@sekula ,@gipezzul ,@lidiaz ,@sutt ,@peter ,@iriu ,@ggonella ,@tamartin ,@lbeemste ,@ademaria ,@suyogs ,@shanisch ,@cantel ,@valentem as watchers
CI Result FAILURE (hash 90faf770) Athena AthAnalysis externals cmake make tests Full details available on this CI monitor view. Check the JIRA CI status board for known problems
Athena: number of compilation errors 0, warnings 0
AthAnalysis: number of compilation errors 0, warnings 0
For experts only: Jenkins output [CI-MERGE-REQUEST-CC7 78169] Hi @rjoshi. From the failing test it looks like this MR changes the trigger counts. Is this intended? If so, can the references be updates?
— James for L1